通过基于图的知识增强提升对话响应生成的事实性
Improving Factuality for Dialogue Response Generation via Graph-Based Knowledge Augmentation
AI总结：
针对大模型对话响应易产生幻觉的问题，提出TG-DRG和GA-DRG两种图知识增强框架，结合推理引导重构、语义知识选择和图增强生成，并引入对话事实分数，在OpendialKG和HybriDialogue上分别提升3.47%和3.12%的事实性。
摘要：
大型语言模型（LLMs）在许多自然语言处理任务中表现出色。然而，它们倾向于产生幻觉——生成看似合理但不一致或事实错误的文本——这可能在特定任务中引发严重问题，包括对话中的响应生成。为缓解这一问题，我们提出了两个新颖的图知识增强框架：基于文本化图的对话响应生成（TG-DRG）和图感知对话响应生成（GA-DRG），它们结合了推理引导的对话重构、对话语义知识选择和图增强的响应生成，以提高对话响应的事实性。为了评估生成响应的事实性，我们提出了一种对话事实分数，该分数解决了现有事实分数方法在对话场景中的局限性，提供了对事实一致性的更可靠评估。我们使用不同的基线方法在OpendialKG和HybriDialogue数据集上评估了我们的方法。与其他图知识增强基线（包括最先进的G-retriever）相比，我们的方法在对话事实分数方面显著提高了事实性，在OpendialKG上提高了3.47%，在HybriDialogue上提高了3.12%。代码将在GitHub上发布。
英文摘要：
Large Language Models (LLMs) succeed in many natural language processing tasks. However, their tendency to hallucinate - generate plausible but inconsistent or factually incorrect text - can cause significant problems in certain tasks, including response generation in dialogue. To mitigate this issue, we propose two novel graph knowledge-augmented frameworks, Dialogue Response Generation via Textualised Graphs (TG-DRG) and Graph-Aware Dialogue Response Generation (GA-DRG), which combine reasoning-guided dialogue reformulation, dialogue sense knowledge selection, and graph-enhanced response generation to improve the factuality of dialogue responses. To evaluate the factuality of generated responses, we propose a dialogue fact score that addresses the limitations of existing fact-score methods in dialogue settings, providing a more reliable assessment of factual consistency. We evaluate our methods using different baselines on the OpendialKG and HybriDialogue datasets. Our methods noticeably improve factuality compared to other graph knowledge-augmentation baselines, including the state-of-the-art G-retriever, achieving improvements of 3.47% on OpendialKG and 3.12% on HybriDialogue in terms of dialogue fact score. The code will be released on GitHub.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
